* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download The Potassium Promoter Function in the
Metastable inner-shell molecular state wikipedia , lookup
Tunable metamaterial wikipedia , lookup
Heat transfer physics wikipedia , lookup
Surface tension wikipedia , lookup
Ultrahydrophobicity wikipedia , lookup
Self-assembled monolayer wikipedia , lookup
Sessile drop technique wikipedia , lookup
Nanochemistry wikipedia , lookup
Low-energy electron diffraction wikipedia , lookup
Langmuir 1993,9, 3427-3440 3421 The Potassium Promoter Function in the Oxidation of Graphite: An Experimental and Theoretical Study Christoph Janiak,tJ Roald Hoffmann,*pt Peter Sjoval1,s and Bengt Kasemo’J Department of Chemistry and Materials Science Center, Cornell University, Ithaca, New York 14853-1301, and Department of Physics, Chalmers University of Technology, S-41296 Goteborg, Sweden Received January 15, 1993. I n Final Form: September 7, 1 9 9 P The interaction of potassium and oxygen, individually and as coadsorbates with graphite, was studied experimentallyand theoretically. In the experiments the sticking of oxygen was investigatedat different K coverages. The extremely low sticking coefficient for chemisorptionldissociation of dioxygen on clean graphite is contrasted by the high sticking coefficient for even very dilute potassium overlayers. KO2 and KzO complexes are identified as early precursors to COZformation from coadsorbed potassium and oxygen. The K-O complexesformed on the surfaceand their transformations (asa function of increasing temperature) were followed by thermal desorptionspectroscopy (TDS)and electron energy loss spectroscopy (HREELS). A theoretical study based on the tight-binding method indicates why the presence of potassium facilitates the chemisorption of dioxygen, whose sticking coefficient on clean graphite is otherwise very small. It is suggested that the potassium promotion effect is a very local one and that formation of a K-02 surface complex is the first step in graphite oxidation. The strongly preferred adsorption site for 02 on a K-covered surface is on top of a potassium atom with no interaction to the graphite layer. From there the dioxygen molecule can dissociate into oxygen atoms. The stabilizingK-02 interaction can be traced to a localized sp-hybrid available for bonding to 0 2 and the fact that the graphite ?r system is not distorted upon chemisorption of 0 2 via the potassium spacer. Concerning the favored adsorption sites for the individual species, a slight preference is deduced for K above a hexagon (6-fold symmetry site), while 0 and 02 more strongly favor a position on top of a carbon atom (02in a bent end-on approach) for the undamaged surface. Surface defects, however, figure prominently in stabilizing the oxygen species and also a potassium atom in the defect area. Introduction Chemical reactions of solid carbon with oxygen, water, and carbon dioxide,known as carbon gasificationreactions, have attracted considerable interest in the last decades, mainly because of their relevance in energy technology.’ It is well established that these reactions are catalyzed by, e.g., alkali salts and transition-metal oxides added to the carbon substrate.2 The reactive properties of solid surfaces are often dramatically modified by adsorption of alkali metals.3 It is a well-known phenomenon that potassium, or more generally, alkali metals can catalyze a myriad of reactions of gaseous substances with surfaces. Experimental observations of this phenomenon are available over the range from semiconductor (e.g. Si, GaAs)4 through semimetal (carbon in the from of graphite or charcoal)”’ to metal318 or metal oxidessurfaces, interacting with dioxygen, water, carbon monoxide, carbon dioxide, etc. Subsequently, this effect is exploited in a wide range of technological + Cornell University. t Present address: Institut fiu Anorganische und Analytische Chemie, Technische UniversitAt Berlin, Strasse des 17 Juni 135, D-10623 Berlin, Germany. Chalmers University of Technology. Abstractpublishedin Advance ACSAbstracts, October 15,1993. (1) Laurendeau, N.M. hog. Energy Combust. Sci. 1978,4,221. (2)McKee, D.W. Chem. Phys. Carbon 1981,16,1. (3)Bonzel, H. P. Surf. Sci. Rep. 1987,8, 43. (4)Starnberg, H. I.; Soukiassian, P.; Bakshi, M. H.; Hurych, Z. Surf. Sci. 1989,224, 13. (5)SjSvall, P.; Hellsing, B.; Keck, K.-E.; Kasemo, B. J. Vac. Sci. Technol. 1987, A5, 1065. (6) SjSvall, P.; Hellsing, B.; Keck, K.-E.; Kasemo, B. Mat. Res. Soc. Symp. Roc. 1988,111,447. (7)(a) Sams, D. A.; Shadman, F. AZChE J . 1986,32,1132. (b) Wen, W.-Y. Catal. Reu.-Sci. Eng. 1980,22,1. (8) Aruga, T.; Murata, Y. hog. Surf. Sci. 1989,31,61. (9) OYoung, C.-L. J. Phys. Chem. 1989,93,2016. @ applications, e.g. in heterogeneous catalysislO and semiconductor oxidation.l1J2 The gasification of coal and charcoal is promoted for over a century by additives such as caustic soda (NaOH) or lime (CaO).2J”16 Enhancement of reactivity by alkali metals has stimulated extensive experimental and also theoretical studies. Various mechanisms have already been proposed for the catalytic effects of alkali additives in the form of alkali metal salts (e.g. carbonates) in the gasification of carbonaceous materials.I6ls The catalytic effect of alkali compounds in carbon gasification reactions is often explained in terms of a mechanism referred to as the oxygen transfer mechanism.2J6 According to this mechanism, the active catalyst undergaes sequential oxidation-reduction cycles on the carbon surface. Oxygen from the gas phase thus is captured on the surface by direct reaction with the catalyst. Subsequent reduction of the catalyst provides oxygen atoms, which react with the carbon surface to form CO and CO2. This mechanism has received experimental support and is usually favored over the so-called electron transfer mechanism.2 In the latter mechanism, the adsorbed alkali is assumed to cause strengthening of (10)(a) Ertl, G.; Weiss, M.;Lee, S. B. Chem.Phys. Lett. 1979,60,391. (b) Campbell, C. T.; Goodman,D. W. Surf. Sci. 1982,123,413. (c) Mroes, W.-D. Catal. Rev.-Sci. Eng. 1983,25,591. (11)(a) Soukiassian, P.;Gentle, T. M.; Bakshi, M. H.; Bommannavar, A. S.; Hurych, Z.Phys. Scr. 1987,35,757. (12)Oellig, E. M.; Michel, E. G.; Asensio, M. C.; Miranda, R. Appl. Phys. Lett. 1987,50,1660. (13)See the special issue of Fuel 1983,62(2),140-262,which contains the papers presented at a symposiumon Fundamentals of Catalytic Coal and Carbon Gasification. (14)Kelemen, S.R.;Freund, H. J. Catal. 1986,102,80. (15)Cabrera, A. L.; Heinemann, H.; Somorjai, G. J. Catal. 1982,75, 7. (16)(a) McKee, D. W. Fuel 1983,62,170. (b) Wood, B. J.; Sancier, K. M. Catal. Reo.-Sci. Eng. 1984,26,233. (17)Mims, C. A.; Pabst, J. K. Fuel 1983,62,176. (18)Moulijn, J.A.; Cerfontain, M. B.; Kapteijn, F. Fuel 1984,63,1043, and references therein. 0743-746319312409-3427$04.00/0 0 1993 American Chemical Society 3428 Langmuir, Vol. 9,No.12, 1993 carbon-oxygen bonds and weakening of the carbon-carbon bonds as a result of electron transfer to the carbon substrate. The influence of alkali metal adsorption on the surface properties of metals, graphite (semimetal), and semiconductors is related to the electropositive nature of the alkali metals which can easily donate their single valence electron to the “surface”. In the physicist’s language, alkali adsorption causes a charge redistribution on the surface that effectivelygives rise to a dipole field which lowers the work function of the s ~ r f a c e , ~The * ~ Jso-called ~ promoting effect of alkali metals in surface reactions is often discussed in terms of this (local) charge redistribution and work function decreaseS2O In a previus experimental study by two of us5s6 we observed that the oxidation rate of evaporated carbon can be greatly enhanced-by up to lo4 times-through a predeposited potassium layer. A theoretical model had also been proposed6 which showed that the large rate increase upon K adsorption could be explained by a K-induced increase in the dissociation rate for the 02 molecules impinging on the surface, due to a more effective coupling between the 02 antibonding T* orbital and the occupied T states in the K-promoted graphite. In other words, a lower work function of K-graphite is proposed to result in more charge/electron transfer to 02, especially its half-filled T* level. This in turn increases the dissociation probability of 02,eventually leading to chemisorption of atomic oxygen.6p21922For metal surfaces a work function decrease is known to affect the adsorption kinetics and the sticking coefficient of dioxygen.3 There is also an “oxygen-transfer mechanism”, the essence of this model being that 02 reacts directly with K and then the KO, complex is reduced by carbon atoms in the surface. Severalchemical forms of the active potassium catalyst have been suggested for the oxygen transfer based reaction mechanisms. The potassium oxides, K20, K202, and KO2, are obvious candidates,because they can undergo oxidation-reduction cycles by transformation between the different oxidation s t a t e ~ . ~Mims J ~ * and ~ ~ Pabst were able to detect the presence of a surface complex on the carbon surface during gasificationconditions.17 Schlogl suggested that the presence of potassium hydroxide, KOH, on the carbon surface is essential for the catalytic effect,24while other reaction schemes include potassium carbonate, K2C03, as an intermediate.2J6 In all these experimental studies, however, the potassium catalyst was added in the form of KOH or K2CO3. Thus, the alkali-catalyzeddioxygen-carbon reaction still presents some unsettled questions: It remains a matter of controversy whether or not the promotion effect is local or n ~ n l o c a l . ~Local ? ~ ~ is here taken to mean that the potassium atom is a direct participant in the reaction and that it undergoes an oxidation-reduction sequence. Nonlocal implies that the increased reactivity of 0 2 is due solely to an electronic influence that the alkali metal has on the carbon substrate. Combinations of these mechanisms are, of course, also possible. (19)Heskett, D.; Maeda Wong, T.; Smith, A. J.; Graham, W. R.; DiNardo, N. J.; Plummer, E. W. J. Vac. Sci. Technol. 1989, B7,915. (20)(a) Nsrskov, J. K.; Holloway, S.; Lang, N. D. Surf. Sci. 1984,137, 65. (b) Lang, N. D.; Holloway,S.; Nsmkov, J. K. Surf. Sci. 1986,150,24. (c) Feibelman, P. J.; Hamann, D. R. Surf. Sci. 1986,149,48. (21)Franciosi, A.; Soukiassian, P.; Philip, P.; Chang, S.; Wall, A.; Raisanen, A.; Troullier, N. Phys. Reu. B 1987,35,910. (22)Hellsing, B. Phys. Rev. B 1989,40,3855. (23)Saber, J. M.; Falconer, J. L.; Brown, L. F. J. Catal. 1984,90,65. (24)SchlBgl,R. In Physics and Chemistryof Alkalt Metal Adsorption; Bonzel, H. P.; Bradshaw, A. M.; Ertl, G., Eds.; Elsevier: Amsterdam, 1989. (25)Albano, E. V. Surf. Sci. 1989,215,333. Janiak et al. Concerningthe mechanism of catalytic carbon oxidation, an important problem is to determine the chemical identity of the potassium catalyst under oxidation conditions. The electron transfer mechanism would be favored if atomic or intercalated potassium were present in the carbon substrate, since these potassium configurations are expected to cause large electron transfer to the carbon substrate.16 Atomic and intercalated potassium are, however, not thermally stable in the graphite substrate at gasificationtemperatures.l4*= Thus, it seems unlikely that these forms of potassium are important for the catalytic effect in carbon gasification. The “potassium-on-graphite” problem is further related to alkali-graphite intercalation compoundsn and to studies of other metal atoms (e.g. Cu, Ag, Au, A1F8 or Ptm) on graphite. The chemisorption of the oxygen molecule (02) on graphite/K is related to the adsorption of 02 on semiconductors4 or metals,30 02 (and also N231) physisorption on graphite,32p33and to the oxygen atom (094 and also N, H,3M7H20,%NO, CO, et^.^^) chemisorption on graphite and other main group elements (semicond u c t o r ~ ) .The ~ ~ distinction between molecular dioxygen ( 0 2 ) and atomic oxygen (0) is significant; the two species differ in their chemical behavior. While atomic oxygen apparently requires very little activation energy to react with a clean graphite surface with a high probability, molecular dioxygen reacts only at high temperature with a much smaller probability and surface defects are said to be important for the reaction^.^^ The dioxygen-graphite surface chemisorption is suggested to be endothermic.341~~ The binding energy of physisorbed 02,on the other hand, is about 0.1 eV, i.e. exothermic. (For the distinction between chemi- and physisorption, see below). The experimental studies referred to above616have been extended to thermal desorption spectroscopy (TDS) (26)SjBvall, P.; Kasemo, B. To be submitted for publication. (27)(a) Riidorff, W.; Schulze, E. 2.Anorg. Allg. Chem. 1964,277,156. (b) Riidorff, W. Adu. Znorg. Chem.Radiochem. 1969,1,223.(c)Inoehita, T.; Nakao, K.; Kamimura, H. J. Phys. SOC.Jpn. 1977,43, 1237. (d) Lagrange, P.; Guerard, D.; Herold, A. Ann. Chim. Fr. 1978,3,143. (e) Dresselhaus, M. S.; Dresselhaus, G. Adu. Phys. 1981, 30, 139. (f) DiVincenzo,D. P.; Rabii, S. Phys. Rev. B 1982,25,4110. (g) Loupias, G.; Codazzi, V.; Rabii, S.; Tatar, R. C.; Ascone, I.; Guerard, D.; Elansari, L. Synth. Met. 1989,34,411.(h) Ohno, T.; Nakao, K.; Kamimura, H.J. Phys. SOC. Jpn. 1979,47,1125. (28)Ganz, E.; Sattler, K.; Clarke, J. Surf. Sci. 1989,219,33. (29)Eppell, S.;Chottiner, G. S.; Scherson, D. A.; Pruett, G. Langmuir 1990,6,1316. (30)Roberta, M. W. Chem. Soc. Rev. 1989,18, 451,and references therein. (31)(a) Wang, S.-K.; Newton, J. C.; Wang, R.; Taub, H.; Dennison, J. R. Phys. Reu. B 1989,39,10331. (b) Barone, V.; Lelj, F.; Ioaconis, E.; Illas, F.; Russo, N. J. Mol. Struct. (Theochem) 1986,139,277. (32)(a) Toney, M. F.; Fain, 5. C., Jr. Phys. Reu. B 1987,36,1248. (b) Mochrie, S.G . J.; Sutton, M.; Akimitau,J.;Birgeneau, R. J.; Horn, P.M.; Dimon, P.; Moncton, D. E. Surf. Sci. 1984,138,599. (c) Toney, M. F.; Diehl, R. D.; Fain, S. C., Jr. Phys. Rev. B 1983,27,6413. (d) Heiney, P. A.; Stephens, P. W.; Mochrie, S. G. J.; Akimitau, J.; Birgeneau, R. J.; Hom,P.M.Surf.Sci.1983,125,539. (e)Pan,R.P.;Etters,R.D.;Kobashi, K.; Chandrasekharan, V. J. Chem. Phys. 1982,77,1035.(0Stoltenberg, J.; Vilches, 0. E. Phys. Rev.B 1980,22,2920. (g) Nielaen, M.; McTague, J. P. Phys. Reu. B 1979,19,3096. (33)Youn, H. S.;Hew, G. B. Phys. Rev. Lett. 1990,64,443. (34)Barone, V.; Lelj, F.; Iaconis, E.; Illas, F.; R w o , N.; Jounou, A. J. Mol. Struct. (Theochem.) 1986,136,313. (35)(a) Mendoza, C.; Ruette, F. Catal. Lett. 1989,3,89.(b) Chen, J. P.;Yang,R.T.Surf.Sci.1989,216,481. (c)Barone,V.;Lelj,F.;Minichino, C.; Russo, N.; Toscano, M. Surf.Sci. 1987,189/190,185. (d) Bennett, A. J.; McCaroll, B.; Messmer, R. P. Phys. Reu. B 1971, 3, 1397. (36)LaFemina, J. P.; Lowe, J. P. J. Am. Chem. SOC. 1986,108,2527. (37)Dovesi, R.; Pisani, C.; Ricca, F.; Roetti, C. J. Chem. Phys. 1976, 65,3075. (38)Miura, K.; Morimoto, T. Langmuir 1988,4,1283. (39)Russo.. N.:, Toscano. M. J. Mol. Struct. (Theochem) 1989.. 201.. 149. (40)Zheng, X.M.; Cao, P. L. Surf. Sci. 1989,219,L643. (41)Wong, C.; Yang, R. T.; Halpem, B. L. J. Chem. Phys. 1983, 78, 3325. (42)Zarif Yanz,Y. A.; Kiselev, V. F.; Lezhnev, N. N.; Nikitina, D. V. Carbon 1967,5,127. Interaction of Potassium and Oxygen studies of K and COZdesorption after initial adsorption of 0 2 , at 136 K, and to vibrational spectroscopic studies of the K,O, complexes formed on the surface. The results most relevant to the present study are presented below. More detailed results are given e l s e ~ h e r e . ~ 5 Previous studies have shown that the probability for dissociative adsorption of 02 molecules impinging on a clean carbon surface, leading to the formation of CO and C02, is very low, at least <103at room temperature, and that the process is thermally a c t i ~ a t e d . ~Furthermore, ~~7 the reaction probability is very sensitive to the surface structure and strongly influenced by defects. For example, the sticking coefficient for 02 on a clean graphite surface when a graphite at 673 K increased from <10-14 to surface with low defect concentration was damaged by Ar+bombardment.4BAt very low temperatures molecular oxygen is physisorbed on graphite. The monolayer desorption temperature is 47 K.48 It seems generally accepted that the sticking and dissociation of dioxygen is frequently the bottleneck of surface oxidation reaction^.^ Note: When referring to the “graphite surface”, we generally mean the basal or 0001 plane of graphite. Experimental Methods and Procedures The experimental details have been described elsewhere.2+s~a In short the experimentswere performed in an ultrahigh vacuum (UHV) system with a base pressure in the low 10-10 Torr range. The system is equipped with a CMA Auger electronspectrometer for surface purity control and for work function measurements, with a high resolution electron energy loss spectrometer (HREELS) for vibrational spectroscopy, and with a mass spectrometer for thermal desorption spectroscopy. Low energy electron diffraction (LEED) is used for surface crystal structure control. The sample was oriented graphite, exposing the basal plane. The sample was cleaned by heating, initially for several hours at 1200 K. Potassiumwas deposited from a break-walglass ampule4 or from an alkali metal getter source.2 Oxygen exposures were normally made in the 10-8Torr range by backfilling the chamber with oxygen. Experimental Results The results are presented in a condensed form with special emphasis on observations of relevance to the theoretical calculations. For additional information see refs 43-45. On the clean graphite surface no oxygen adsorption (sticking coefficient <103-10-4) was observed down to 136 K, and no reaction to form CO or COZwas observed in 0 2 atmospheres up to 106 Torr and 700 K. This finding is consistent with earlier work,40?41@@ indicating no or very weak dioxygen chemisorption and a large activation barrier for 0 2 dissociation. In contrast the 0 2 sticking coefficient on potassiumcovered graphite is unity or close to unity, even at very dilute potassium c0verages.~3The initial sticking coefficient is constantlynear unity, at least down to a potassium coverage of 5% of a full monolayer (Figure 1). This behavior can be e ~ p l a i n e deither ~ ? ~ ~in terms of a direct interaction of 0 2 with the graphite surface, modified by electron donation from adsorbed potassium, or as a (43) Sj6vall, P.; Kaeemo, B. J. Chem. Phys. 1993, $8,5932. (44) Sj6vd, P.; Kaeemo, B. Surf. Sci., in press. (45)(a)Chakarov, D.;Sjhall, P.;Kasemo,B.J.Phys.: Condenu.Matter 1993, 5, 2903. (b) Chakarov, D.; bsterlund, L.; Kasemo, B. To be submitted for publication. (c) Zhdanov, V. P.; Kasemo, B. Submitted for publication in Chem. Phys. (46)Barber, M.; Evans, E. L.; Thomas,J. M. Chem. Phys. Lett. 1973, 18, 423. (47) Marchon,B.;Carrazza,J.; Heinemann, H.; Somorjai,G. A. Carbon 1988,26, 507. (48) You, H.; Fain, S. C., Jr. Phys. Rev. B 1986,33, 5886. Langmuir, Vol.9, No.12,1993 3429 t 0 o*5! a .Y .0 ti 0 0 0 0.5 Potassium coverage 1 Figure 1. Sticking coefficient of 0 2 on potassium-covered graphite as a function of potassium coverage. precursor adsorption4k of 0 2 into the physisorbed state,M where it eventually reacts with potassium atoms on the surface. In the former case the mechanism is a direct charge transfer to the 0 2 molecule impinging on the surface, enhanced by the lowered work function and possibly also by the increased electron density at the Fermi leveL6 In the second case the reaction occurs as a direct reaction between a weakly perturbed, physisorbed 02molecule and potassium atoms on the graphite surface. The theoretical analysis below will elucidate these questions. As a consequence of 02 sticking on the potassiumcovered surface, various K-0 complexes are formed, as revealed by the HREELS spectra (Figure 2A, for details see ref 45a,b). At least three different species are observed with vibrational energies around 27, 31, and 39-41 meV. The 31-meVpeak is not obvious in Figure 2A, the spectrum at 160 K, but clearly identifiable at different exposure and coverage conditions.46b A t low temperatures, the vibrational loss peaks are fairly broad, indicating heterogeneous broadening (Le. there is a mixture of local stoichiometries due to kinetic barriers). As the structures formed at low temperature are annealed, the peaks sharpen and undergo transformations. Some of these transformations are accompanied by desorption of potassium and/or CO2 (Figure2B). An additional fador influencing the surface K/O stoichiometry is potassium accompanyingpotassium deposition and annealing of the different K-0 coadsorbate layers. By variation of the surface stoichiometry (i.e. the potassium/ oxygen ratio), and guided by recent XPS/HREELS results the loss peak at 39 from 02 adsorption on meV is assigned to KzO (for the detailed assignments, see ref 45a). The loss peak at 31meV is especially pronounced at oxygen-rich conditions and is attributed to KO2 or possibly K202. The loss peak at 27 meV appears only after annealing of the K-0 coadsorbates to 700 K, which then causes considerable desorption of potassium (Figure 2B). A definitive assignment is yet lacking. It is interesting to note that this is the only loss peak remaining after annealing to 700 K. Combining this observation with the CO2 desorption occurring around 750 K (Figure 2B)leads to the conclusion that the 27-meV loss peak represents a precursor to C02 formation. It is tempting to speculate about the possibility of a carbonate species, but the lack of evidence for a C-O stretch vibrationaldoes not at present provide any support for this hypothesis. (A C-0 bond parallel to the surface would, however, escape detection by dipole scattering, so the existence of a CO bond cannot be excluded.) (49) Baddorf, A. P.; Itchkawitz, B. S. Surf. Sci. 1992,264,73. 3430 Langmuir, Vol. 9, No. 12,1993 A HREELS: ( 0 2 + K ) cnGraphite ? a .~ X1a: 0 27 50 100 150 Energy Loss (mew Janiak et al. I* B TDS: (02+ K ) on Graphite 350 ak 550 650 Temperature (K) 450 750 Figure 2. (A) HREEL spectra after 25 langmuirs 02 exposure of 2.5 monolayers (ML)potassium on graphite and after annealing the surface to the indicated temperatures in vacuum. All spectra are recorded at 160 K. Primary electron energy 15 eV, specular scattering. The elastic peak shown corresponds to the unannealed system. (B)Thermal desorption spectra of potassium, m/e = 39, and carbon dioxide, m/e = 44, (a) after 1.2 ML deposition of potassium at 160 K on graphite, and after 3 (b), 8 (c),and 25 langmuirs (d) oxygen exposures at 160 K. Linear heating rate of 2.5 K d . The spectra for COZare shifted upward for the sake of clarity. The COz spectrum (a) is multiplied by a factor of 3. The arrows correspond to the annealing temperatures in Figure 2A. The thermal desorption spectra (Figure 2B) contain desorption peaks for K and COZ. The K peak around 450 K corresponds to metallic (unreacted) potassium, the peak at 500-550 K is due to potassium stabilized by oxygen, and the peak around 750-800 K is coninciding with COZ desorption. By combining TDS and HREELS results, the following picture emerges. At low temperatures oxygen sticks on the potassium-covered surface with unity or nearly unity sticking even at low K coverages. At low potassium coverage the average K:O ratio at saturation is about 1:1, as judged from the sticking and TDS experiments, but due to intercalation the surface ratio is probably smaller. At high K coverage the saturation ratio is 2:1, indicating KzO formation, and in agreement with the HREELS data. At low temperatures, kinetic barriers give rise to a mixture of stoichiometries, Upon annealing, the surface K-0 complexes go through a number of transformations. At temperatures up to 300400 K, segregation of intercalated potassium increases the surface K:O ratio. At even higher temperatures some potassium desorbs from the surface, and as a result more oxygen-rich complexes are formed. Eventually total decomposition/reaction occurs to form COz and desorbing K. The precursor species to COZformation is associated with the 27-meV vibrational energy,rabut its composition is yet unidentified. In mixed oxygen isotope experimants complete isotopic mixture is observed, among product CO;! molecules,44indicating complete dissociationof dioxygen.47 The observations above point to the key role of dissociation of oxygen and potassium bonding to graphite and of K-0 and C-0 bonds on the surface. These are central questions in the theoretical analysis below. Theoretical Preparations The extensive experimental research in the area of graphite-adsorbate interactions has also generated a wealth of theoretical studies. Most of the calculations do not employ an extended, two-dimensional surface, but use cluster models instead.36B8 To our best knowledge the actual chemisorption of 0 2 on graphite, however, has only been studied once by theoretical methods. We want to contribute a chemcial perspective to this problem. Our calculationsmake use of the tight-binding extended Huckel method on an infinite, two-dimensional surface.60 Before we can embark on a full scale calculation, some decisions need to be made. A One- or Two-Layer Slab? In many studies the graphite surface is simulated with a planar C,H, cluster model, an annulene system. In tight-binding calculations one graphite layer is typically u~ed.3S-3~t~~ A one-layer annulene or graphite-net system is reasonable, since the interlayer distance in graphite of 335 pm is close to a van der Waals contact. The effect of going from a one-layer to a two-layer slab is for the most part negligible, as our calculations show.62 However, the introduction of a second layer leads to an inequivalency of the two carbon atoms within the surface unit cell. One carbon atom has a carbon directly underneath in the next layer (asite), whereas the other one is located above the center of a carbon hexagon in the next layer (/3 site)-illustrated in 1. 1 In an STM (scanning tunneling microscopy) picture of the basal plane of graphite, three carbon atoms of the hexagon appear much brighter than the other three, with the bright and not so bright spots alternating.2816-6 The brightest spots were assigned to the /3 sites, since calcu(50) Hoffmann, R. Solid and Surfaces: A Chemist’s View of Bonding in Extended Structures, VCH New York, 1988. (51) (a) Perkins, P. G.; Marwaha, A. K.; Stewart, J. J. P. Theor. Chim. Acto 1980,57,1. (b) Messmer, R. P.; McCarroll, B.; Singal, C. M. J. Voc. Sci. Technol. 1972, 9, 891. (52) Note, however, that two vs one graphite layer makes a difference in the total (integrated) DOS at the Fermi level. The one layer model results in a so-called zero band gap semiconductor, i.e. zero state density at q,whilethe three-dimensionalcalculationsyield a small overlap between the valence and conduction bands, resulting in a nonzero DOS at q. (53) Wiesendanger, R.; Anselmetti, D.; Geiser, V.; Hidber, H. R.; Giintherodt, H.-J. Synth. Met. 1989, 34, 175. eM- Interaction of Potassium and Oxygen .. Langmuir, Vol. 9,No. 12,1993 3431 -r -5 0-site a-site -8 > - & f - E f -7 ," -11 ' x P E -13 w _- I -- I -17 -20 r M K I1 r DOS DOS X a b Figure 3. Partial band structure diagram (a) of a two-layer graphite slab. Density of state (DOS) plots (b) for a small region around K (denoted by X, within dashed lines in a, or see circled area with enlargementin the drawing of the Brillouin zone, above). The DOS plots show the total DOS (dotted lines) and projected DOS for the a- and @-sitecarbons (shaded area). 81 k points were used for this average property calculation. lations showed these to have a higher electronic density of states (DOS)near the Fermi level-which is observed by STM-than the a sites.However, recent STM studies of intercalated graphites57 and single graphitic sheetson metal surfaces%argue against this interpretation. Also, it could not be completely ruled out that other effects might contribute to these huge STM corrugations, such as surface layer slippage and elastic deformations induced by the close proximity of the tunneling tip.59 The origin of the unusual graphite STM images is still a matter of controversy;54+60,61 one of us has recently done a thorough study of the problem, suggesting the possibility of charge and spin density waveP playing a role in the observed graphite STM images. With our calculations we do not see any difference in the density of states for a and 6 sites when we average over the whole Brillouin zone. Only if we concentrate on the fraction of the surface Brillouin zone near the Fermi level, near special k-point K,% (illustrated in 2) can we discriminate between the two different carbon atoms. u + = k-points have the higher DOS right at the Fermi level-in agreement with the other computational results mentioned above. Focusing on the region around K in our calculation is necessary since not all k points in the Brillouin zone contribute to the charge densities sensed by STM-only k points around the Fermi level do. For typical tunneling voltages below 1 VM the portion of the band structure sampled by STM near the Fermi surface is very small; the Fermi level lies close to the special point K. For a simple, pedagogically effective approach to the ?r-band crystal orbitals and their degeneracy lifting in two- and threedimensional graphites near the Fermi energy; see ref 62. To return to the question posed at the beginning of this section, we conclude that it would be fairly safe to use only one graphite layer to model the surface. The inequivalencies between the two different carbon sites do not show up in a two-layer slab, neither on a bare surface nor with adsorbates attached to it. However,a two-layer slab allows for more options for simulation of defects (holes) in the surface which are important for the reactivity of graphite. Therefore, we opted for a two-layer slab. Choice of Unit Cell for Chemisorption Studies. Potassium-promoted oxidation of graphite using pure potassium has been observed and studied experimentally for potassium submonolayerto monolayer coveragesst6(see experimental section). We need a unit cell large enough to treat submonolayer coverages. This is mostly the range explored experimentally and also the one of primary interest for the theoretical work. Larger coverages are expected to represent metallic potassium rather than potassium on graphite. Actually alkali-metal overlayers are frequently reported to become metallic somewhere above 0.1 monolayer c0verage,6~,~ reflecting the large effective radius of the alkali metal.6 For potassium, the minimum metal-metal separation in its elemental body . ~ ~ where centered cubic structure is 454.4 ~ p m Exactly the metallic character of the alkali overlayer is well established can be debated and most likely depends on the substrate. For example, the plasmon frequency and work function characteristicof metallic K is not established until coverages higher than 0.1 ML, typically around 0.5 ML. We tried to simulate a low-enough coverage throughout our calculations by selecting a large unit cell (cf. ref 36; concept of a molecular unit cells1). Our choice was a unit cell of 18 carbon atoms per layer (or 36 C's for the two layer slab) as depicted in 3. The area spanned by this unit cell is about 4.7 x 10-15 cm2. With an approximated monolayer coverage of 5 X 1014 K-atoms per cm2,3one alkali-metal per unit cell then corresponds to a -40 % or 0.4 monolayer coverage. Note again, that here a monolayer or fraction thereof is referred to a close packed potassium layer and has no correspondenceto the graphite structure, due to the incommensurate nature of the carbon and potassium atomic radii. 2 Figure 3 shows the DOS plots for the a and sites from a small region around K. We see that the 6 sites clearly (54) Batra, I. P.; Garcia,N.; Rohrer, H.; Salemink,H.; Stoll,E.; Ciraci, S. Surf. Sci. 1987, 181, 126. (55)Tomhek, D.; Louie, S. G.; Mamin, H. J.; Abraham, D. W.; Thomson, R. E.; Ganz, E.; Clarke, J. Phys. Rev. B 1987,35,7790. (56) Tomhek, D.; Louie, S. G. Phys. Rev. B 1988,37,8327. (57) (a) Kelty, S. P.; Lieber, C. M. J.Phys. Chem. 1989,93,5983. (b) Kelty, S. P.; Lieber, C. M. Phys. Rev. B 1989,40, 5856. (58) J. C. Hemminger, to be submitted for publication. (59) Tersoff, J.; Lang, N. D. Phys. Rev. Lett. 1990, 65, 1132. (60) Lawunmi, D.; Payne, M. C. J. Phys.: Condens. Matter 1990,2, 3811. (61) Tchougreeff, A. L.; Hoffmann, R. J. Phys. Chem. 1993,97,8993. , 3 We are now ready to embark on our actual surfaceadsorbate calculations, which might provide us with an ~~ (62) LaFemina,J. P.; Lowe, J. P. Int. J. Quantum Chem. 1986,30,769. Janiak et ai. 3432 Langmuir, Vol. 9,No. 12,1993 understanding of the alkali-promoted oxidation of graphite. Specifically, we would like to find the preferred adsorption site for potassium and to understand the reason for the low sticking coefficient of dioxygen on a bare graphite surface, so as to comprehend the role of the potassium catalyst in enhancing the sticking and 0 2 dissociation. The latter should help us to comment on the local versus nonlocal controversy discussed in the introduction and suggest some possible reaction mechanisms. Finally, we also want to explore how defects in the basal plane of graphite may influence its reactivity toward oxygen. The Preferred Potassium Adsorption Site. On a graphite slab one can have four higher symmetry sites for a monoatomic adsorbate: (A) the “%fold on-top site” directly above a carbon atom. Since the carbon atom can be an a or /3 carbon, as described earlier, we can, in principle, distinguish between an Aa or an A/3 site. One may also envisage (B)the “2-fold bridging site” above the center of a carbon-carbon bond and (C)the “6-foldhollow site” above the center of a hexagon. The different sites are illustrated in 4. 4 In this section we use an approximate molecular orbitd method, the extended Hiickel tight-binding procedure. This one-electron method gives only qualitative results but lends itself to the construction of simple explanations. Quantitiescomputed by the extended Huckel methodology should be viewed as relative indications of stability of bonding. In particular, the extended Huckel method is not expected to give reliable results as far as bond length variations are concerned. Therefore, we chose a C-K distance of 303 pm, based on structural studies of K-graphite intercalation compounds27(305 pm found), [r15-(PhCH2)5C5]K(thf)366* (average 303.5 pm; thf = tetrahydrofuran), [ v ~ - ( M ~ ~ S ~ ) C(average ~ H ~ ]300 K ~pm), ~ (average 303 pm). This and [~~-Me5C5]K(pyridine)2~~~ C-K distance was kept constant throughout our studies. We only note that in potassium-fluorenyl systems67the K-C distances are longer (range 307-332 pm) and vary more than in potassium-cyclopentadienyl complexes66 (maximal variation there 296-310 pm). Table I compares the binding energy, the charge on K, and the C-K overlap population for the four adsorption sites given in 4. As expected (from what is known of graphite-potassium intercalation compounds and molecular cyclopentadienylsystems),the $4 site (6-fold hollow) is the most stable one for the potassium adsorbate. It not only has the highest binding energy but also gives the largest potassium-graphite overlap population, the latter (63) Gaspar, J. A.; Eguiluz, A. G.; Tsuei, K.-D.; Plummer,E. W. Phys. Rev. Lett. 1991,67,2854. Liebsch, A. Phys. Rev. Lett. 1991,6?,2858,and references therein. (64) Frank, K.-H.; Sagner, H.-J.; Heskett, D. Phys. Rev. B 1989, 40, 2767. (65) Wyckoff, R. W. G. In C r y s t a l S t r u c t u r e s , 2nd ed.; Wiley-Interscience: New York, 1963; Vol. 1, p 14. Table I. Comparison between Different Potassium Adsorption Sites on Graphite As Illustrated in 4 C-K overlap Eb,binding charge on sitea enermlevb potassium ACX AB 6.14 6.14 6.18 6.40 +0.62 +0.62 +0.81 +0.83 B C populationC 0.055 0.055 0.049 (X2) 0.032 (X6) a See 4 for illustration; C-K distance fixed at 303 pm. Eb = E(bare graphite unit cell) + E(free K) -E(surface + adsorbate) = (-2556.708 eV) + (-4.34 eV) - E(surface + adsorbate), positive is surfaceadsorbate complex more stable; Note that & is defined relative to a free K 4s energy of -4.34 eV. The singleabsolute value Of&?A depends critically on the potassium parameters (see Appendix), so that only the trend noted here is of physical significance. Overlap population to one carbon at a distance of 303 pm from the potassium. For the B site there are two; for C there are six carbons at this distance. To obtain a relative measure of the K-graphite bond strength, the C-K overlap population should be multiplied by this number of neighboring carbons as indicated in parentheses. reflecting qualitatively the bond strength. In this hollow position the potassium can also transfer its electron most effectively to the graphite surface. The difference in the Fermi energy between the different adsorption sites (not shown in the table) is negligible. The small variation in binding energies for different adsorption site may indicate that it is easy for the potassium to move around on the graphite surface. The near identity of the Aa! and p sites is no surprise, after what we saw for the a and p sites for a bare grahite surface. We note that this finding of a 6-fold hollow sitepreference for K is different than what has been observed experimentally for the coinage metals and aluminum. The coinage metals-Cu, Ag, and Au-have a similar valence configuration,namely dl0s1,to the alkali metals. However, STM studies show they prefer A and B sites.% Chemically, from a comparison with cyclopentadienyl model compounds, this does not come as a surprise: Cp-CUI, -Agl, or -Aul complexes exhibit a a-bonded, ql-Cp-metal coordination, rather than a R-, $-bonding mode. What we did not look at was the stability of the 6-fold potassium adsorption site with respect to alkali metal clustering. A strong tendency toward formation of metal adlayer islands at low coverages (about 1%of a metal monolayer) has been observed experimentally in STM studies of Cu, Ag, Au, and A1.28 However, the potassium adsorption is accompanied by a larger charge transfer. Coulomb repulsion would then tend to keep the positively polarized K adsorbates apart at low coverages. The Preferred Dioxygen Site on Bare Graphite. There is much geometrical freedom in the way diatomic 0 2 can interact with the graphite surface. Relative to the surface plane the 0-0bond vector might be (1)“side-on” (5), aligned parallel to the surface plane, (2) linear “endon” (6), perpendicular to the surface plane, or (3) bent “end-on”(7),the vector forming an angle y with the normal to the surface plane. “Side-on”and “linear, end-on” are, of course, specialpositions (y = 90 and 180’) of the general case shown in 7. O--0 P /”Y 0 ) 0 I 77777777777777775 6 7 In experimental studies of monomeric dioxygen metal complexes, a linear or close to linear “end-on”coordination Langmuir, Vol. 9, No. 12,1993 3433 Interaction of Potassium and Oxygen of dioxygen has not yet been f0und.a If 0 2 is not side-on bonded to a metal center, it assumes a bent end-on position with y typically in the range between 115 and 135’. Therefore, we limited out theoretical studies also to the side-on and bent end-on cases. The higher-symmetrical adsorption sites which we studied are illustrated in 8 for the side-on and in 9 for the end-on, mode. an optimized C-O bond and not to a C-0 distance for chemisorbed 0 2 . 1 , 2 - D i o ~ e t a n e s(10) ~ ~ ~or~ peroxiraness9t71(11) can be considered as molecular models for our 02-graphite surface study. Depending on reaction path and method, a C-O separation of 176 and 200 pm, respectively, has been calculated for transition states for the ethylene42 reaction.@ P 10 11 Subsequent studies showed us that we can take the region between 175 and 200 pm as appropriate for the C-0 distance for 0 2 chemisorbed on the basal graphite plane. (We use the word “chemisorbed” to indicate the 8 regime of bond distances, where there is considerable C-0 interaction, but do not mean to indicate that there necessarily is a chemisorbed state.) Table I1 compares the binding energy and 0-0overlap populations for sideon chemisorbed dioxygen on graphite in different sites, assuming C-O distances of 150, 175, and 200 pm. The calculated binding energies are all negative, indicating the endothermic character of the interaction. As 9 anticipated, the Aa and j3 sites differ only very slightly. site) seems to be Having 0-0 lying above a C-C bond (BII A refers again to on-top sites (aand p), B to sites over the most likely side-on position. On the other hand, C the center of a C-C bond, and C to sites over the center sites and BI are highly unfavorable. In all cases the 0-0 of a hexagon. Within these three basic sites one can overlap population is lowered, compared to the free 0 2 descriminate further among various conformations. Howmolecule, due to a fillingof the dioxygen A* level. A smaller ever, we elaborated on such conformational changes only overlap population should be accompanied by a lengthfor the side-on case. The most important difference occurs ening of the 0-0bond. The effect of C-0 distance is such in the B site; we can have 0-0aligned parallel to a C-C that at a longer separation of, e.g. 200 pm, the effects bond (eclipsed, or Bll conformation) or perpendicular to discussed here are essentially the same, albeit less proit (staggered, BJ. For the C site we can make a similar For instance, the binding energy is less negative. nounced. distinction (CIIor CI,see 8). The conformational differThe increasingly negative binding energies a t a shorter ence may be expected to be less pronounced in A and C C-O distance of 150pm underscore the 2-orbitaV4-electron compared to B sites, for the rotational barrier in A and type repulsive interaction between 0 2 and the graphite C is 6-fold. surface. Instead of focusing on different conformations for the The respective values for the bent end-on approach in bent end-on mode we carried out a small angle-variation different sites and with different surface-0-0 angles are study (y = 150, 135, and 120O) with the second oxygen compiled in Table 111. While the C sites are again highly approaching the surface as shown in 9. disfavored, the A position starts to become slightly The 0-0 distance throughout most of our study was exothermic. At a longer C-0 separation (200 pm, results fixed at 121 pm-the separation in free 0 2 . While 0 2 not reported here) A is even more exothermic and B also adsorption will surely result in lengthening of the 0-0 shows a positive binding energy. The angle variation is bond, the reason for one site being preferred over another not very conclusive, ita trend depending largely on the should emerge more clearly if the 0-0 distance is kept adsorption site. Again, for the bent end-on chemisorption, fixed. 0 2 a* level lies below the Fermi energy. the The choice of a suitable C-0 distance is more probThe general trend that the binding energy becomes lematic. In the absence of experimental studies of positive at larger distancesfrom the surface is in agreement chemisorbed 0 2 on graphite, we were left with theoretical with the readily observed physisorption of dioxygen on studies of the process. We are aware of only one graphite, with surface-adsorbate distances of the order of investigation of chemisorption of 0 2 on a CloHs (naph300 pm.% On the other hand, dioxygen chemisorption on thalene) molecule in which C-0 distances were optimized. graphite is mostly an endothermic process. A bent endDifferent methods give a C-0 distance in the range of on approach of the dioxygen molecule in the on-top site 129-142 pm for the chemisorption product (cf. Table 3 in (A) is calculated as most favorable for real chemisorption. ref 34). However, we feel that these values correspond to Here we note that graphite studies of 0 2 indicate that (66)(a)Lorberth,J.;Shin,S.-H.;Wocadlo,S.;Masaa,W.Angew.Chem.there may not be such a thing as a chemisorbed 0 2 , but only physisorbed 02,32148since to our knowledge no 1989,101,793;Angew. Chem.,Int.Ed.Engl. 1989,28,735. (b) Jutzi, P.; Leffers, W.; Hampel, B.; Pohl, S.; Saak,W. Angew. Chem. 1987,99,663; molecular state except the physisorbed 0 2 has ever been Angew. Chem., Znt. Ed. Engl. 1987,26,583. (c) Rabe, G.; Roesky, H. W.; observed experimentally. We want to emphasize that the Stalke, D.; Pauer, F.;Sheldrick, G. M. J . Organomet. Chem. 1991,403, 11. (67) (a) Janiak, C. Chem. Ber. 1993,126,1603. (b) Zerger, R.; Rhine, W.; Stucky, G. D. J. Am. Chem. SOC.1974,96,5441. (68) (a) Boca, R. Coord. Chem. Reu. 1983,50,1. (b) Gubelmann, M. H.;Williams,A.F.Struct.Bonding(Berlin)1983,55,1. (c)Niederhoffer, E. C.; Timmons, J. H.; Martell, A. E.Chem. Rev. 1984,84,137. (69)Hotokka, M.; Roos, B.; Siegbahn, P. J. Am. Chem. SOC.1983,105, 5263. (70)Kopecky, K. R. In ChemicalandBiological GenerationojExcited States; Adam, W., Ed.;Academic Press: New York, 1982; p 86. (71)Chen, C.-C.; Fox, M. A. J. Comput. Chem. 1983,4,488. 3434 Langmuir, Vol. 9, No. 12, 1993 Janiak et al. Table 11. Comparison between Different Sites for Side-on Chemisorbed Dioxygen on Graphite as Illustrated in 8 at C-0 = 150, 176, and 200 pm overlap population ~~ site’ Aa AB 150 -7.82 -7.81 -4.15 -13.6 -10.5 -11.8 175 -3.89 -3.88 -2.06 -6.49 -5.88 -7.04 C-Od 0-0‘ Eb, binding energy/eVb 200 -1.23 -1.22 -0.17 -2.59 -1.84 -2.68 150 0.492 0.491 0.531 0.407 0.627 0.644 200 150 175 200 0.470 0.104 0.045 -0.014 0.468 0.104 0.045 -0.014 0.488 0.404 0.233 0.026 BII 0.457 0.085 0.031 -0,001 0.551 0.337 0.201 0.045 CI 0.580 0.323 0.167 0.043 0 See 8 for illustration, 0-0 was fixed at 121 pm. Eb = E(bare graphite unit cell) + E(free 02)- E(adduct) = (-2556.708 eV) + (-249.681 eV) - E(adduct). A positive binding energy indicates stabilization. The overlap population in free 02 with 0-0 = 121 pm is calculated as 0.793. Overlap population between one carbon and one oxygen a t the distance given. For the A, Bll, and CIIsites there are two, for BI and CL there are four C-0 contacts to consider for a relative measure of the Orgraphite bond strength. Table 111. Comparison between Different Sites and Bending Angles for Bent End-on Chemisorbed Dioxygen on Graphite As Illustrated in 9 at C-0 = 175 pm site’ Ae B C 8urface-Q-O angle, y (deg) 150 135 120 160 135 120 150 135 E,,, binding energy/eVb 0.143 0.245 0.113 -0.400 -0.442 -0.833 -10.30 -11.10 overlap PoP‘Jlation C-od 0.462 0.238 0.481 0.253 0.495 0.266 0.468 0.193 0.480 0.190 0.487 0.181 0.530 -0.007 0.527 -0.033 175 0.469 0.469 0.493 0.447 0.597 0.614 of carbon atoms and requires only the formation of fourcoordinated carbon atoms in reasonable geometries. The A site for the side-on mode presents also an interaction with one carbon atom only, but it makes this carbon atom five-coordinated. If we allow for a synergistic simple distortion or reconstruction around the interacting carbon atom(s),such that the atoms involved are slightly elevated above the graphite plane (12, 131, the 02 binding energies increase somewhat, even after correcting from the reconstruction energy. An elevation of d = 50 pm above the surface corresponds to almost tetrahedral angles around carbon. 0 See 9 for illustration, 0-0 was fixed at 121 pm. b-d See notes on Table 11. e Because of the close similarity between a and B sites, no distinction is made anymore. 12 E(graphite-0,) 13 thermic 1I 11 t -200 pm -300 pm d(graphite-02) chemisorption physisorption region Figure 4. Schematic representation of the energetics of t h e graphite-dioxygen interaction. Physisorption is intended to mean a van der Waals interaction while chemisorption should describe a chemical (orbital) interaction. Physisorption energies are generally in t h e 0.1-eV range, also for 02 on graphite.43 Approximate distances are from ref 34. results clearly demonstrate that there is no stable state for 02 chemisorption. At most, the molecule may only be trapped in a weakly bound state for a limited time. In this respect 02-graphite is very much like H2 on CU’~ and H2 on Mg.73 The van der Waals interaction is probably the dominant attractive term for 0 2 interacting with clean graphite. We definephysisorption as a mere van der Waals interaction, while chemisorption in our nomenclature is intended to mean a chemical (orbital) interaction, i.e. a shift in electron density. Figure 4 summarizes what we extract as the common, accepted picture concerning the energetics of 0 2 chemi- and physisorption on graphite. The preference for the A site for bent end-on or the BII site for side-on bonding derives, we think, from the fact that in these positions 0 2 interacts with the least number (72)Harris,J.; Anderason, S . Phys. Reu. Lett. 1985,55, 1583. (73)Narskov, J. K.;Haumeller; Johansson, P. K.; Lundquist, B. I. Phys. Rev. Lett. 1981,46, 257. For a small reconstruction (d = 25 pm) the gain in binding energy is largest (0.6 eV more for A, bent end-on; 1 eV more for BII, when compared to the undistorted case). The experimentally observed low sticking coefficient for 02 (at temperatures higher than 50 K)32p48suggests that the resulting energy for the distored A site is still not large enough to trap an 02molecule effectivelyin a chemisorbed state (because kT is too large in comparison with the well depth). Potassium facilitates the distortion only (and to a very limited amount) if the elevated carbon atom is part of the hexagon above which the alkali metal sits. With potassium close by, the energy required for the distortion decreases by about 0.1 (d = 25 pm) to some 0.4 eV (d = 50 pm). Oxygen Atoms on Graphite. The experimental findings point to a difference in reactivity between a dioxygen molecule and an oxygen atom. We need to examine theoretically oxygen atoms on a graphite surface, to check if pronounced differences do indeed exist. We start again with a comparison between different sites (cf. 9) for chemisorbed monooxygen on graphite. The results are compiled in Table IV. For the A and B site we find rather large positive binding energies (indicatingstabilization) with the on-top site being slightly favored over the bridging site. The C-site or 6-fold hollow position can be excluded immediately. The stabilizing chemisorption energies for an 0-atom at a C-0 distance used before for dioxygen correlate with our expectations. It makes chemical sense that one obtains a “stable” chemisorption adduct from a highly reactive species such as monooxygen. Langmuir, Vol. 9, No. 12, 1993 3435 Interaction of Potassium and Oxygen We can further justify the formationof gaseousoxidation products (CO, CO2) from the monooxygen chemisorption, since the C-C overlap population of the adjacent bonds drops significantly from 1.016 for clean graphite to 0.877 for 0 in the on-top site. This difference in C-C overlap population becomes even larger if we distort the interacting carbon by elevating it 25 or 50 pm above the surface (cf. 12). With 0 on-top, the C-C overlap population to its neighbors decreases to 0.33 in both cases-down from 0.91 (50 pm above surface) to 0.99 (25 pm elevation) for the clean distorted graphite. At the same time the binding energy for monooxygen increases. We may add that experimentally it would be of interest to explore the branching ratio CO:CO2, when only atomic oxygen is present on the surface. As far as we know, these ratios 02:CO:CO2 for atomic oxygen (no potassium) on the surfacehave not been measured. With K on the surface there is only negligible CO and 0 2 desorption.4 It would be of interest to deposit atomic oxygen, followed by measurement of the 02:CO:CO2 ratios. Reactivity of Defects in the Graphite Surface. Microscopic techniques such as STW4and etch-decoration transmission electron microscopy as well as scanning electron mi~roscopy4l,~~ show clearly the formation of monolayer pits on graphite upon its gasification reaction with atomic and molecular oxygen. It is proposed that these pits start at existing defects (e.g. atomic vacancies) and can be readily expanded with atomic oxygen at about 150 “C at a rate which is several times faster than the oxidation rate in molecular oxygen at 650 Although some expansion of the pits with 0 2 , after preceding 0 atom exposure, is apparently possible, its extent is not clear. Work on graphite gasification by molecular oxygen is usually carried out at elevated temperatures (650 “C and above), where 0 2 can thermally dissociate. It is quite possible that at these temperatures sufficient atomic oxygen is formed to dominate the reaction.41 In our own experimental work, we also observed a reaction-inducedetch pit at the center of one sample that had been used extensively in reactive experiments. We suspectthat a cooperative effect of the electron beam (with energy up to 10 keV), and reactions involving coadsorbed K and 02,was responsible for the pit formation. We tested the influence of an atomic defect, i.e. a C vacancy on the basal graphite surface, by first removing a single carbon atom (from an a site) and then adsorbing a dioxygen molecule (bent end-on), or an oxygen atom, above the vacancy (14). /O 14 The binding energy for a vacancy with 0 adsorbed right above it indicated only a minor improvement, while for 0 2 it was even less favorable compared to the defect-free surface. The 0-0and C-0 overlap populations also do not give a clear indication whether a single atom vacancy will facilitate the sticking and dissociation of an 02 or 0 adsorbate. Such a small effect on a surface defect worried us at first, especially so in light of recent STM studies that indicate the ready formation of larger pits on preexisting (74) Chang, H.; Bard, A. J. J. Am. Chem. SOC.1990,112,4598. (75) Walker, P.L.,Jr. Carbon 1990,28, 261. Table IV. Comparison between Different Adsorption Sites for Monooxvgen on Undistorted GraDhite at C-0 = 175 pm overlappopulation site EhIeVa C-O C-Cd& Ab +3.10 0.231 0.877 0.187 0.756 (bond underneath 0) B +2.28 C aEb -8.68 -0.007 0.912 + = E(graphite unit cell) E(free 0)- E(adduct). A positive bindingenergyindicatesan exothermicadsorptionprocess. Becaw of the close similarity between a and B sites, no distinction is made between these. defects by heating highly pyrolytic graphite samples in air at 650 OC.76 We then tried to derive an understanding of the frontier orbitals-the danglingbonds-with the use of a molecular model. A look at the molecular orbitals of 15 in comparison to 16 showed that the unsaturated c 16 15 levels are raised in energy-but so are the R levels. There is a concentration of electron density on the carbenoid carbons, but c is still substantially below the R level (see 17). /17 The symmetry plane of the molecule prevents mixing between c and R dangling bonds, so that there is little electron density in the region above the vacancy. However, things -are slightly more complicated, as shown by a theoretical study which was carried out by Soto on graphite (one or two layer slab, extended Hiickel tight-binding method) with a singleatom vacancy.77 The results indciate a distance dependence of the charge density directly above the vacancy. This is explained by the localized nature of the nonbonding crystal orbitals from the nearest neighbor atoms to the vacancy. In the calculated STM image the vacancy would appear as a “hole” only at low altitudes, whereas it would look like a maximum or “bump” at high altitudes, with the changeover occurring at an altitude of about 175 pm. From the orbital pictures in 17 we could expect a more pronounced defect influence if we had the 02 or O-adsorbate approach the defect site either above the rim or in the surface plane-as indicated in 18. nm f O’? 18 We thus explored putting dioxygen and monooxygen on top of a rim atom (19) and-after removing a second (76) Chang, H.; Bard, A. J. J. Am. Chem. SOC.1991,113,5588. (77) Soto, M. R. J. Microsc. 1988,152, 779. 3436 Langmuir, Vol. 9, No. 12,1993 Janiak et al. carbon atom-in the plane of the surface layer (20). Adsorbing 0 or 0 2 in the plane of the graphite surface with one vacancy only would require C-0 distances of 142pm! We therefore decided to remove a second carbon atom which createsa large enough vacancy to accommodate an 0 atom with C-0 = 175 pm. patterns, which differ in their K-02 separation and relative orientation (taking into account the potassium atoms in the neighboring unit cells). The same is true for dioxy- 0do’ I 19 20 Note that the in-plane approach, modeled in 20,results in simultaneous interactions with four carbon atoms with danglingbonds. For the rim position, (19)we find binding energies which are about 1.5 eV more positive than the respective values for the undamaged graphite surface. For the in-plane approach 20 the binding energies are smaller by more than 2 eV, which we, however, attribute to the square-planar environment for the oxygen atom. Actually, we would expect a larger positive energy of chemisorption for the in-plane trajectory than for the rimsite, because of the localized nature of the dangling cr orbitals (versus ?r) and their better energy match with the (di)oxygen levels. Therefore, to model a more reliable in-plane approach, we had to extend the pit even further. This could only be done by resorting to an even larger unit cell of 24 carbon atoms for the surface layer (plus 32atoms for an eventual second layer), shown in 21. 22 gencoadsorbed in thy bent end-on position above a C-atom (A site). With the freedom of 0-0 rotating about the C-0 contact, we even have various conformations in addition to three different K/O2 chemisorption patterns (23). However, for simplicity we calculated only three conformers with the second O-atom above a C-C bond, as shown in 23. 23 21 Indeed, for this large enough defect we find in-plane binding energies which are an additional 2-3 eV more positive than for the respective rim-position. We carried out these calculations using a one- (defect surface only) and two-layer (see 21) slab. From our computational parameters we fail to see any effect on the second layer, with respect to which dioxygen is then in the physisorption range. Experimentally, the physisorption binding energy is found to be rather small, only -0.1-0.2 eV.43 We note that surface defects (one to multiatom vacancies, 14-21) also have a stabilizing effect for potassium atoms. The binding energies for potassium adsorbed on top of a rim atom (cf. 19),or over a vacant site (cf. 14),or in an in-plane position (cf. 21)such that C-K = 303 ppm, are some 0.2-0.5 eV stronger than for respective positions on the undamaged surface. This is consistent with the large increase in desorption temperature for potassium from an argon bombarded (and thus damaged) surface.44 Potassium-Dioxygen Coadsorption. We have now set the stage for an investigation of potassium-dioxygen coadsorption on a graphite surface. There now enters a new parameter, the K-02 separation, in addition to the already mentioned different possible adsorption sites for potassium and dioxygen. The schematic drawing in 22 illustrates what is meant; having potassium in a 6-fold hollow (C) site and dioxygen coadsorbed,e.g. above a C-C bond (Bll site),allowsfor four different K-02 chemisorption Furthermore, we can envision an entirely new chemisorption site for the oxygen molecule when it is brought onto the potassium/graphite surface: 0 2 can sit above a preadsorbed K atom, either side-on (24)or bent end-on bonded (25). These adsorption sites will be denoted K1 and K2,respectively. -I 1 I I b K1 site 24 K2 site 25 For these sites the K-O distance was fixed at 270 pm, based on a comparison to the inorganic complexes pm), K[Cs(C02Me)& K [ C S ( N O ~ ) ~(K-O C ~ ~= ] ~258-294 ~ (MeOH)79(K-0 = 264-288 pm), and [(PhCH&CsIK( O C ~ H B(K-0 ) ~ ~ 269-278 pm). In Table V we compare different potassium-dioxygen chemisorption patterns with respect to the dioxygen binding energies, 0-0and K-0 overlap populations. The potassium position was fixed in the C site (K-C = 303pm, (78) Otaka, Y.;Marumo, F.;Saito, Y. Acta Crystallogr. B 1972,28, 1590. (79) Bruce, M. I.; Walton, J. K.;Williams, M.L.;Hall, S.R.;Skelton, B. W.;White, A. H.J. Chem. Soc., Dalton Tiam. 1982,2209. Interaction of Potassium and Oxygen Table V. Comparison between Different Potassium-Dioxygen Chemisorption Patterns on Graphite at C-0 = 175 pm K-0 E ~binding , overlap population 02 siteo distanceshm enerat/evb o-Cle K-O -- BII 1 Bl12 B113 B114 A1 A2 A3 K1 K2 Side-on, See 22 0.501 0.009 3851385 -1.66 0.504 0.095/4.009 2971374 -1.32 0.521 0.346/4.052 1671282 +0.18 0.543 0.209 1531153 +1.20 End-on, See 23 (135' C-0-0tilt angle) 387/369 +1.14 0.479 0.002/0.022 299/335 +1.38 0.486 0.07810.008 170/228 +2.88 0.522 0.230/0.054 Above Potassium Atom, See 24 and 25 2701270 +6.35 0.483 0.075 2701366 +6.13 0.465 0.193/-0.026 0-0was fixed at 121pm. Eb = E(carbons + potassium) +E(free 02)- E(surface1adsorbate complex). A positive binding energy indicates an exothermicadsorption process. c The overlap population in free 02 with 0-0 = 121 pm is calculated as 0.793. see above). We include only dioxygen sites (except for the novel K sites) which were found to be most likely from our Oz/graphite calculations, namely the BI site for sideon and the A site for end-on chemisorbed b 2 . The sites deemed unlikely from our earlier calculations remain so, even with potassium as a coadsorbate. For dioxygen in the BIIand A sites we note the strong shift of the binding energy toward more positive values, indicating the increased exothermic character of the 0 2 adsorption on a graphite surface with a preadsorbed potassium layer. In other words, the energy lowering upon impingement of an 0 2 molecule on the surface (as one condition for sticking) is enhanced drastically by a potassium coadsorbate. The closer the oxygen approaches a potassium atom, the stronger the binding energy becomes. The large binding energies for the K sites (dioxygen on top of potassium) are striking. They surpass the comparable values for all other chemisorption patterns by more than 3 eV. Moreover, the stabilization provided by potassium for the 0 2 chemisorption (in the K site) is still some 1.5-2 eV higher than the already sizable stabilization derived from defects in the graphite surface (in-plane approach of 02; 21, E b = 4.10 eV end-on; 4.64 eV side-on). In the following we try to rationalize the local nature of the potassium enhancement with the help of molecular models. But before we do so, we would like to show that at our level of computation, molecular models preserve the trends seen from the investigation ,of extended structures. Table VI compiles some molecular systems and compares binding energies and overlap populations to the values in analogous extended structures. Even though Table VI presents a very brief list, one realizes that the overlappopulations tend to be surprisingly similar in molecular and extended systems. The binding energy values may vary a little bit, but the trend in stabilization upon going from side-on bonded 0 2 , over bent end-on, to 0 2 above a potassium, is clearly preserved. If we now examine the interaction diagrams between 0 2 and the molecular C, fragments (not shown here), we note some subtle differences in the orbital interactions: For 0 2 side-on bonded to C16Hlo we have a very strong perturbation of the carbon a system through a repulsive 0 2 a*/C a interaction. Such a repulsive interaction gives rise to the destabilization of side-on adsorbed 0 2 on graphite (see above). For 0 2 end-on bound to C16H10, on the other hand, the repulsive 0 2 a*/C 7r interaction is rather weak and a 2-orbital/2-electron 0 2 a/C, a interaction becomes predominant with a smaller carbon a-level perturbation. Langmuir, Vol. 9,No. 12,1993 3437 When 0 2 is adsorbed on top of a potassium, sitting in turn above a C6H6 ring, one sees only a very small perturbation for the benzenoid a system and the 0 2 a* level-illustrated in Figure 5. We see again an important role of the 0 2 a and 0 2 a orbital as they interact with the potassium s-p hybrid. It is this hybrid pointing away from the surface which mediates the sticking and chemical interaction of 0 2 with the graphite surface at a C-0 distance which is normally in the physisorption region. The large positive binding energy for 0 2 on top of the potassium can be traced to the transfer of the remaining electron density on potassium into the 0 2 a* level.5 Dioxygen Dissociation. It is generally accepted that the gasification of graphite is preceded by chemisorption (sticking)of the dioxygen molecule on the surface, followed by its dissociation to oxygen atoms. In our investigation we emphasized the (dissociative) sticking component as the apparent bottleneck in the reaction process. And indeed, as the 0-0 overlap population in all of the tables shows, the 0-0 bond strength is substantially lowered when the dioxygen molecule interacts with the surface-even in cases with a negative binding energy. From much experience a lowered overlap population correlates with a weaker bond. Thus, a lengthening of the 0-0 bond should occur. This may eventually lead to bond dissociation. Let us explore this dissociation in some more detail. Figure 6 shows the Walsh diagram (orbital energies versus atomic separation) for the 0-0 bond stretch in a free dioxygen molecule. The left-hand value on the abscissa-121 pm-is the equilibrium distance in free 0 2 . The doubly degenerate a* level is half filled with two electrons, giving rise to the well-known paramagnetic character of dioxygen in its triplet 32g-.ground state. The lowering of the overlap population upon chemisorption stems from the filling of the antibonding a*level. The oxygen a* (at about -13 eV) lies below the Fermi level of the graphite slab, around -11 eV. One has, of course, to be careful here since in our simple one-electron approximation the 0 2 a* would be filled even at large separation from the surface. However, if one is in a region where adsorbate-substrate interaction is likely to occur (cf. C-0 = 200 or 175 pm), we may rely on these results. Putting two more electrons into an antibonding orbital will, of course,weaken the corresponding bond. Occupying 0z2-)is not yet, by itself, sufficient the a* level in 0 2 to break the bond. With the overlap population dropping from 0.79 to ca. 0.46-0.50 (cf. Tables 11-VI) the atoms can still be consided to be held together by the a bond. However, from molecular dioxygen-metal complexes it is known that as the number of antibonding electrons rises from 0 2 through 0 2 - to 0 2 2 - the 0-0 distance increases from 1 2 1 over 133 to 149 pmm (accompanied by a distinct lowering of the 0-0 stretching frequency). As pointed out above, such a bond lengthening-possibly to approximately 150 pm, depending on the adsorption site-will also occur with dioxygen on the graphite surface. Returning now to Figure 6 we realize that the strongly antibonding a* level experiences a dramatic decrease in energy upon lengthening of the 0-0bond. Our qualitative calculations show that the energy of a* approaches the Fermi level of the surface at an 0-0 separation of about 150-160 pm. Once it sinks below the Fermi energy (tf) it will be filled by two electrons from the slab (26), giving rise to a dissociation of the molecule into oxygen atoms. An additional experiment to test for chemisorbed active monooxygen may be the use of carbon monoxide as a reducing material. A recent study indicated that the (4 (80)Vaska, L.Acc. Chem. Res. 1976, 9, 175. 3438 Langmuir, Vol. 9, No. 12,1993 Janiak et al. Table VI. Comparison between Molecular Surface-Adsorbate Models and the Analogous Extended Structure System molecular system overlap population 0-0 , , 1 , 0/" 1 analogous extended structure overlap population Eb" 0-0 c-0 Eb' -2.18 0.49 0.24 -2.06 0-0 0.49 0.23 (cf. Table 11, BII) -0.08 0.48 0.26 +0.25 0.48 0.25 (cf. Table 111, A, 135') -0.06 0.48 0.26 +0.25 0.48 0.25 0.87 +3.10 0.88 (cf. Table IV) 0.87 +3.10 0.88 C-O 135" H H ) T/" 13Y H "I. H H H H H H H +2.75 0 H +2.77 0 H H molecular system analogous extended structure overlap population Q>,? overlap population Eb" 0-0 K-O 0.51 K-O 0.08 +6.35 0.48 0.08 (cf. Table V,Kl) 0.48 0.08 +6.35 0.48 0.08 Eb" 0-0 +5.00 +5.94 , , , / I I I , K l -+H H H 9-? -" H Eb H H 'H = E(substrate) + E(free 0/02) - E(adduct). A positive binding energy indicates an exothermic adsorption process. oxidation of CO may be used as a probe reaction to detect some forms of chemisorbed oxygen in coals.81 '4 - I I 26 We have focused here on the 0 2 dissociation step, since it seems to be the more important rate limiting step in CO and C02 formation of graphite. We did not explore in any detail the C-0 bond formation nor the detachment of. CO/C02 from the surface. Also left for a later study are (81)Hinckley, C.C.;Wiltowski,T.;Wiltoweka, T.; Ellison,D. W.; Shiley, R. H.; Wu, L. Fuel 1990, 69, 103. the formations of K-0 and K-C-0 complexes on the surfaces and their likely stoichiometries. Theoretical Conclusions Supplementing the experimental results, we have tried to provide a possible understanding of parts of a graphite gasifcation mechanism, especiallythe dioxygen adsorption and dissociation. We attempted to work out the differences between 0 2 adsorption on clean or potassium-covered graphite in a chemical, molecular orbital picture. We found that in the absence of potassium on the surface there is no deep potential energy well in 0 2 chemisorption, allowing only physisorption of 0 2 molecules with low kinetic energy.32 A deep potential minimum developes only for oxygen atoms. However, experimentally this requires substantial kinetic energy input in the appropriate degrees of freedom (translational, vibrational and/or rotational energy), either by raising the temperature or providing the 0 2 with kinetic energy using molecular beam techniques to overcome the 0 2 2 0 dissociation barrier. With potassium on the surface this dissociation barrier is - Interaction of Potassium and Oxygen 0-0 \ / K I I C6H6 i x* 0 m-3 _--- . 02 K I C6H6 Figure 5. Partial interaction diagram for dioxygen bonded sideon to a benzenepotassium fragment (K1site). Lack of perturbation of the benzenoid u system together with a good overlap between the dioxygen Q and ?r level and a potassium sp hybrid explains the large energy of interaction. A vertical bar indicates the occupation of the HOMO. -3 Langmuir, Vol. 9, No. 12, 1993 3439 Acknowledgment. C.J. thanks the Deutsche Forschungsgemeinschaft (DFG) for the award of a postdoctoral and a Habilitation fellowship. The theoretical work carried out by C.J. initiated during a postdoctoral stay at Cornell in the group of R.H. and was finished upon return to Berlin. At Cornell our work was supported by the Office of Naval Research and the National Science Foundation through the Materials Science Center at Cornell, Grant DMR-9121654. The TU Berlin is thanked for providing computer time, the “Fonds der chemischen Industrie”, and the “Freunde der TU” for additional support. We thank John Lowe for helpful suggestions and for bringing some of the theoretical work on graphite to our attention and one of the reviewers for his or her very constructive comments. Furthermore, we are grateful to Dinko Chakarov for valuable discussions and for help with the HREELS measurements. Appendix The computations were performed within the extended Hiickel formalismB2with weightedHijvalues.% The atomic parameters for the elements involved in our calculations were as (Hii,0: C 29, -21.4 eV, 1.625; 2p, -11.4 eV, 1.625;82 0 2s, -32.3 eV, 2.275; 2p, -14.8 eV, 2.275;82 K 49, -10.49 eV, 1.2; 4p, -7.37 eV, 1.2; from charge iteration, see text. The parameters for the charge iteration of the potassium atom were A = 1.00304/1.34340; B = 6.63321/4.49320; C = 4.30229/2.72900. Geometrical parameters and k-point sets are referred to in the appropriate tables or in the text. The intralayer C-C distance in the undistorted graphite net was fixed at 142 pm, the interlayer contact at 335 pm. Chosing the Brillouin Zone and k-Point Set. Our bare graphite slab in 3 still has 3-fold symmetry. Together with the inversion symmetry of the Hiickel Hamiltonian the irreducible part of the Brillouin zone is l/12 of the full zone. However, bringing in an adsorbate, such as potassium or 0 2 , above a C-C bond, already reduces the symmetry to 2-fold, with the irreducible Brillouin zone (IBZ)then occupyingone-quarter of the full first Brillouin zone (27).37 Coadsorption of K and 02 (28) may leave no symmetry at all, thus giving only an oblique coadsorbate/ substrate unit cell. , -15 120 130 140 150 160 170 0-0distance / pm bl Figure 6. Partial Walsh diagram for the stretching of the 0-0 bond in the dioxygen molecule. 121 pm is the medium distance in free 02. The doubly degenerate u* level is the half-filled, highest occupied MO. significantlyreduced due to the destabilization of the 0-0 bond, caused by charge transfer from potassium to antibonding 0 2 orbitals. We suggest the following “dynamic” picture of 0 2 approaching K-graphite: As 0 2 approaches a K atom, it first fills 7r* which leads to a bond length increase, which in turn shifts c* down until it falls below Ef. Then c* becomes populated and the 0-0bond weakens drastically until dissociationoccurs. We see the potassium promotion effect as a very local one, in that the alkali strongly facilitates the chemisorption of dioxygen with formation of a K-02 surface complex. The promoter effect of potassium surpassesthat of surface defectsand distortions. The potassium catalyst’s role can be described as enabling the stickingof 02and subsequent dissociationon graphite at a surface-02 distance where normally only physisorption occurs. A potassium s-p hybrid pointing away from the surface-together with no endothermic disturbance in the graphite 7r system-provides for the stabilizingK-02 interaction. For atomic monooxygen we find the expected stabilizing chemisorption already with no potassium present. IBZ b2 21 b.c I IBZ 28 For the special case of an oblique lattice with ~alag= 60’ and Lblb2 = 120°,respectively,the unit cell in reciprocal 3440 Langmuir, VoZ.9,No. 12,1993 Janiak et al. space of the obliquelattice correspondsto the full Brillouin zone (BZ) of the hexagonal lattice, as depicted in 29. The shaded or lined areas of the oblique BZ can be mapped onto an area of the hexagonal lattice by means of an Therefore, there is no problem in inversion center (0). using the k-point sets for an oblique lattice also in cases of higher hexagonal symmetry. Actually, even in 6-fold symmetrical hexagonal lattices it is safer to average over the whole BZ (shown in 29) instead of just the irreducible BZ (given in 2), to ensure a reliable calculation of properties. bl 29 bi 30 In order to be consistent in average property calculations, and to be able to compare results for different systems, one has to use the same k-point set for both low- and high-symmetry systems. Due to computational constraints, we decided on using a grid with only eight k-points-though a highly efficient onea-shown in 30 (the k-points are indicated by + - labels), throughout our average property calculations. The efficiencyof the small k-point set in 30 was verified through a comparative average property calculation on graphite using 18 and 72 k-point sets. The Potassium Parameters. The parameters of the extended Huckel method are the Coulomb integrals Hii and Slater orbital exponents {. The choice of these for systems spanning a range of ionic and covalent bonding is always a problem. For carbon and oxygen there exist well-established literature parameters. However, taking a set of potassium parameters for neutral K, which we found in the literature (H48= -4.34 eV, H4p = -2.73 eV, {48,4p = 0.8785)creates serious problems, as it gives rise to strong counterintuitivemixing (COM)in our semiempirical calculations. COM has been described before, and the conditions for its occurrence were analyzed in detai1.83986 We just mention briefly the features of this phenomenon in our problem. The effect is easiest illustrated by using a molecular model for our extended surface system. (82) (a) Hoffmann, R. J. Chem. Phys. 1963,39,1397. (b) Hoffmann, R.; Lipscomb, W. N. J . Chem. Phys. 1962,36,2179; 1962,37,2872. (c) Whangbo, M.-H.;Hoffmann, R.; Woodward, R. B. h o c . R. SOC.London, Ser. A 1979,366, 23. (83) Ammeter, J. H.; Btirgi, H.-B.; Thibeault, J. C.; Hoffmann, R. J. Am. Chem. SOC.1978,100, 3686. (84) Ramirez, R.; Biihm, M. C. Int. J. Quantum Chem. 1988,34,571. (85) Alvarez, S.; Mota, F.; Novoa, I. J. Am. Chem. SOC.1987,109,6586. (86) Whangbo, M.-H.; Hoffmann, R. J. Chem. Phys. 1978,68,5498. Y r x 31 If we carry out a molecular calculation on a neutral benzene-potassium model (31) using the potassium parameters given above and a reasonable C-K distance of 303 pm (see above), we compute an atomic charge on potassium of +1.21(!), a K-C overlap population of -0.08, and gross atomic orbital populations for the potassium s, y, and x/z orbitals of -0.05, -0.13, and -0.01, respectively (despite using the weighted Wolfsberg-Helmholtz formulas3). A calculation with potassium on the extended graphite surface gives very similar values. These unrealistic values can be attributed to the above mentioned counterintuitive orbital mixing-the out-of-phase admixture of an energetically high-lying metal orbital into a very low lying ligand combination.83~86 In our case the main interaction responsible for COMtakes place between the potassium py and the benzene all-bonding, a1 carbon 2s combination-graphically illustrated in 32 (the numbers denote the orbital energies in eV). To a smaller extent the potassium s orbital is also involved. -2.7 -4.3 +, '\ '\ -29.9 ''\\A -29*5 K K I C6H0 C6H6 32 The potassium y and s orbitals overlap substantially with the C6H6 a1 combination (overlap matrix element -0.56 for y and 0.40 for s), while they differ much in ionization potentials (see 32). These are the conditions leading to COM within the extended Hackel framework. To obtain a better set of parameters, we therefore performed a charge iteration on potassium on an extended graphite surface. For the charge iteration the potassiumgraphite geometry was kept constant with C-K = 303 pm and potassium placed above a 6-fold site at 0.4 ML coverage (1K/18C atoms). A one-layer graphite surface had to be used in the charge iteration procedure, due to computational constraints. In addition we carried out < charge iterations for different exponents (0.87 < 1.40) to find an optimum parameter set which would give a small positive C-K overlap population and a potassium charge as close to +1 as possible, yet still having diffuse potassium orbitals ({48,p as small as possible). The set with an exponent {a,p= 1.20 with H48 = -10.49 and H4p = -7.37 eV was then our final choice and the basis for the subsequent calculations involving potassium on graphite.